Support NeoGAF

[Durante]

Welcome to unreleased software/specifications?

Normally, when a new multi-vendor API is introduced, the specification is released and open long before the first implementation. For OpenCL it was almost a year. New OpenMP additions are often published in papers 2 or 3 years before being adopted in the standard.

Mantle certainly doesn't follow any of the common patterns of open standard development.

 

[iamshadowlark]

Normally, when a new multi-vendor API is introduced, the specification is released and open long before the first implementation. For OpenCL it was almost a year. New OpenMP additions are often published in papers 2 or 3 years before being adopted in the standard.

Mantle certainly doesn't follow any of the common patterns of open standard development.

There aren't multiple vendors, yet. The plan is to bring more people into the fold. And they've already said that the collaborative effort that comes may not be "Mantle" at all. It doesn't have to follow some cookie cutter scheme to be open.

Can we create official MANTLE |OT| ? Switching between 4 different threads is confusing and i am sure as hell new threads about Mantle will pop out every few days.

go for it.

 

[Perkel]

Can we create official MANTLE |OT| ? Switching between 4 different threads is confusing and i am sure as hell new threads about Mantle will pop out every few days.

 

[Dictator93]

Can we create official MANTLE |OT| ? Switching between 4 different threads is confusing and i am sure as hell new threads about Mantle will pop out every few days.

Please. Please. Please

I am so confused by this "multi-threading."

lol

 

[Turin Turambar]

People have no patience. They want things now.

As Durante said, other standards needed years of meetings, specifications, unification, etc.

Right now, Mantle isn't open, true.

First, I would say they wanted to get out what we have right now, a proof of concept done with a few developers to raiser public awareness and make a point of how a low level API is something good on pc.

Later this year, once they have a body of evidence of the improvements and some public supporting them, they will post specifications and initiate talks with other bodies of the industry. Then maybe in two years we will have a Open Mantle standard.

They are being smart, I think this way they have improved their chances of success. First start it privately to make a point and show your willingness to the project, and then initiate the talks. The other way, the companies could have been going on in talks that wouldn't go anywhere and ambiguous statements about maybe yes, maybe no.

 

[Durante]

There aren't multiple vendors, yet. The plan is to bring more people into the fold. And they've already said that the collaborative effort that comes may not be "Mantle" at all. It doesn't have to follow some cookie cutter scheme to be open.

Those "cookie cutter schemes" are what has been adopted after literally decades of experience with open standards, because they work.

I wish I could believe that AMD is really only into this to improve 3D graphics for everyone, but if that was the case, and if CPU overhead is that important to them, then why did it take them years to support multithreaded rendering / deferred contexts / display lists in DirectX11, and why does their support still frankly suck on some of their cards?

 

[alexandros]

At the very least everyone should give AMD the benefit of the doubt. At least they said they'll try and turn Mantle into an open API, which is more than one can say about Nvidia and their own proprietary technologies.

 

[KojiKnight]

They are being smart, I think this way they have improved their chances of success. First start it privately to make a point and show your willingness to the project, and then initiate the talks. The other way, the companies could have been going on in talks that wouldn't go anywhere and ambiguous statements about maybe yes, maybe no.

This. AMD has tried on MANY occasions to go the other way to create standards and the results have been very mixed. This is a different approach to be sure, but I agree it's a good one. Instead of "hey everyone, let's get together and maybe fix perceived problems in DX!" it's "This is what we can do to make it better. Everyone can try it themselves and see! Now let's get together and make this work for everyone."

As for it becoming a standard? I think even if it never becomes an "open" standard, a standard it's likely to become. It's not easy to convince several big developers to jump onto a new technology, but the fact that one of the biggest games already HAS and a promise of every new Frostbite engine game also supporting it already makes it more successful than many other standards.

 

[Durante]

At the very least everyone should give AMD the benefit of the doubt. At least they said they'll try and turn Mantle into an open API, which is more than one can say about Nvidia and their own proprietary technologies.

If they truly do work to make it open (realistically, and that crucially includes giving up control), then all power to them.

What really rubs me the wrong way is all this marketing of something as open that couldn't be more closed right now.

 

[Perkel]

So who is doing Mantle OT ? I can't obviously...

 

[iamshadowlark]

Those "cookie cutter schemes" are what has been adopted after literally decades of experience with open standards, because they work.

I wish I could believe that AMD is really only into this to improve 3D graphics for everyone, but if that was the case, and if CPU overhead is that important to them, then why did it take them years to support multithreaded rendering / deferred contexts / display lists in DirectX11, and why does their support still frankly suck on some of their cards?

You really need to ask this question? Is it not blatantly clear which approach they've chosen in regards to supporting multi-threading?

So who is doing Mantle OT ? I can't obviously...

I'll do it.

 

[KojiKnight]

So who is doing Mantle OT ? I can't obviously...

Good point I'll start gathering the required info if you guys can come up with a decent enough title for me.

 

[wildfire]

thanks for answer mate. im running a OC i7 4770 (@4.4) so no real benefit in either case for me ill just enjoy what they can use in the PS4 with mantle

I forgot to make this disclaimer. Just wait until we see multi GPU results. There may be even bigger performance gains than single GPU.

 

[Durante]

You really need to ask this question? Is it not blatantly clear which approach they've chosen in regards to supporting multi-threading?

So you both applaud their decision not to support the standard mutlithreading features of DirectX11 well, and admit that Mantle development is to the detriment of their driver quality in standard APIs?
Just for the record.

 

[KojiKnight]

Okay, got the OT thread made, just need a witty title!

Thinking something simple "AMD Mantle OT - Time to shake some things up!"

(edit) Meh, just went with it. I'm creatively bankrupt but I'm also lazy

OT Here http://www.neogaf.com/forum/showthread.php?p=99317150#post99317150

 

[iamshadowlark]

So you both applaud their decision not to support the standard mutlithreading features of DirectX11 well, and admit that Mantle development is to the detriment of their driver quality in standard APIs?
Just for the record.

I didn't applaud anything.

 

[tarheel91]

Star Swarm Stress Test


Its purpose is different to Benchmark and it even doesn't say benchmark in name. Steam called in benchmark in shop but it isn't as name suggest.

-Benchmark is created as linear as possible to get every bench as close as possible.
-Stress test is created to get out of your PC last juices increasing amount of things on screen and so on.


It is good to compare total power of PC with amount of ships on screen, batches and so forth. FPS is secondary in this situation. And if you want to use FPS to compare you need to stick it with other numbers from demo.

Just because it's not consistent doesn't mean the data can't be used. The variance is effectively random and will be balanced out with a decently large sample size.

 

[Raydeen]

Forgive my ignorance, but I remember plenty of OpenGL games back in the day...just how did Direct X win out in the end if it was so inferior?

 

[alexandros]

Forgive my ignorance, but I remember plenty of OpenGL games back in the day...just how did Direct X win out in the end if it was so inferior?

 

[DieH@rd]

Forgive my ignorance, but I remember plenty of OpenGL games back in the day...just how did Direct X win out in the end if it was so inferior?

Way better dev tools, and MS invested in DX development for a while.

 

[wildfire]

Forgive my ignorance, but I remember plenty of OpenGL games back in the day...just how did Direct X win out in the end if it was so inferior?

There were tradeoffs and in the end DirectX won.

Here is a good historical analysis.

http://programmers.stackexchange.com/questions/60544/why-do-game-developers-prefer-windows

Many of the answers here are really, really good. But the OpenGL and Direct3D (D3D) issue should probably be addressed. And that requires... a history lesson.

Birth of Conflict

One day, sometime in the early 90's, Microsoft looked around. They saw the SNES and Sega Genesis being awesome, running lots of action games and such. And they saw DOS. Developers coded DOS games like console games: direct to the metal. Unlike consoles however, where a developer who made an SNES game knew what hardware the user would have, DOS developers had to write for multiple possible configurations. And this is rather harder than it sounds.


OpenGL Ascendant

Thus the stage is set: Direct3D vs. OpenGL. It's really an amazing story, considering how bad D3D v3 was.

Dawn of Shaders, Twilight of OpenGL

Then, GeForce 3 came out. And a lot of things happened at the same time.

One Language to Ruin Them All

So, the OpenGL development environment was fractured for a time. No cross-hardware shaders, no cross-hardware GPU vertex storage, while D3D users enjoyed both. Could it get worse?


Falling Towards Apotheosis

While I maintain that 3D Labs struck the fatal blow, it was the ARB itself who would drive the last nail in the coffin.

The actual article is 8 word document pages long, before reading all the linked citations.

 

[Raydeen]

There were tradeoffs and in the end DirectX won.

Here is a good historical analysis.

http://programmers.stackexchange.com/questions/60544/why-do-game-developers-prefer-windows



The actual article is 8 word document pages long, before reading all the linked citations.

Thanks. Reading now.

 

[Rafterman]

I really don't see why this constantly get brought out as a conversation point. It makes me question just how ept someone is in understanding low-level programming.

What the API is designed for is irrelevant, an API is only a list of pre-fabbed commands one can call to supplement functions. There's no heuristical boundries for NV/Intel/Whoever to jump over in supporting this, if it came to fleshing out their own products. There's no dependencies they would have to AMD's commands.

The only thing that has to be common is syntax, which begets no real advantage to either party.

If the API is designed to take advantage of certain features and advantages of AMD cards it most certainly matters what the API was designed for. And you don't know anything about what dependencies there are or how gimped Mantle might be for another vendor. I think the idea that AMD secretly developed their own API with the intention that it would boost performance on Nvidia cards equally as well as their own is naive as hell.

If you don't think there will be an advantage for AMD over Nvidia using the same API that was designed in house by AMD you're drinking the Kool-Aid. You give them way more credit than they deserve at this point, given the fact that they haven't even released how this API will even work with other vendors and they haven't reached out to them in the entire time the API was being developed.

 

[Reckoner]

On my 7970 I get better performance on Battlefield 4. On Siege of Shanghai it stays at 60fps almost all the time (Ultra settings, ofc). I say almost because there are some strange dips to 45/50 fps. Something really sudden and kind of annoying for half a second.

I know it is still being optimized for my card, so I'll keep with Dx11 until they release a more reliable update.

 

[coughlanio]

On my 7970 I get better performance on Battlefield 4. On Siege of Shanghai it stays at 60fps almost all the time (Ultra settings, ofc). I say almost because there are some strange dips to 45/50 fps. Something really sudden and kind of annoying for half a second.

I know it is still being optimized for my card, so I'll keep with Dx11 until they release a more reliable update.

I'm seeing the same on my 290X.

Mostly constant 60FPS on Ultra, but with sharp sudden drops to 45 for a split second.

 

[artist]

If you don't think there will be an advantage for AMD over Nvidia using the same API that was designed in house by AMD you're drinking the Kool-Aid. You give them way more credit than they deserve at this point, given the fact that they haven't even released how this API will even work with other vendors and they haven't reached out to them in the entire time the API was being developed.

lol, people at this point must be thinking that AMD is a charity of a non-profit effort ..

Remember folks, it was the developers that approached the big three (Intel, Nvidia & AMD) for lower level access and it was only AMD that took it on (again for obvious reasons). I assume AMD besides providing lower level access, also provided quite a bit of resources (support) to repi and his team initially and then when they were up and running more devs like Nixxes etc came into the fold.

From what repi said that Mantle so far hasn't used any or GCN's proprietary stuff which is why it "could" work on Nvidia GPUs as well.

 

[Telephoto999]

I'm seeing the same on my 290X.

Mostly constant 60FPS on Ultra, but with sharp sudden drops to 45 for a split second.

Are you running vsync?

Try gametime.variablefps 60 instead

 

[iamshadowlark]

If the API is designed to take advantage of certain features and advantages of AMD cards it most certainly matters what the API was designed for. And you don't know anything about what dependencies there are or how gimped Mantle might be for another vendor. I think the idea that AMD secretly developed their own API with the intention that it would boost performance on Nvidia cards equally as well as their own is naive as hell. If you don't think there will be an advantage for AMD over Nvidia using the same API that was designed in house by AMD you're drinking the Kool-Aid.



.

I don't know how I can explain this further.

It doesn't matter what functions AMD has layed out in the API, Nvidia wouldn't be able to use them in most anyway. Why? Because they are two completely different architechures. Again, its low-level programming. By that distinction, the fallacy that Nvidia or any other IHV is forever bound to whatever disadvantages and predispositions that you so surely believe AMD has booby trapped for them is just that, a fallacy. Again, one of the few things they could share is syntax and other minor normalites that provide no advantages to anybody.

There's no Kool-Aid to drink, only an understanding of computer science which you seem to lack.

. You give them way more credit than they deserve at this point, given the fact that they haven't even released how this API will even work with other vendors and they haven't reached out to them in the entire time the API was being developed

How do you know this again?

 

[dr_rus]

No... FUD spreading is when you make bold statements that go against what a company says without any proof such as saying "spreading AMD lies".

People asked about Mantle being an open standard... AMD has stated that LATER THIS YEAR (which is hardly "eventually") they plan to bring more companies in to work on it and to try and make it an open standard (or something very like Mantle, meaning AMD is flexible enough that if enough changes need to be made they'd as to make it a different product, they'd STILL adopt it over Mantle because they want a new API standard over D3D)

What reason do you have to believe that AMD is lying about this statement? Where is your proof? You have some right? If you don't then you are spreading FUD! If you DO then please share so the rest of us can see "AMD's lies".

You can't make an open standard from a standard that was made as a closed one, developed in secrecy for two years without any input from other vendors. All the talk from AMD about how they plan to make it open and bring in more companies is lying - because if they'd plan to do this they would've done this when they started working on Mantle and not after they released API, drivers and games supporting it. Right now there is almost zero chance for any support for Mantle from Intel, MS and NV. They simply won't support something that was designed without any input from them to make their products (CPUs, Windows/DirectX and GPUs) look worse. So who will they bring? What "companies"? How will they make Mantle relevant outside of AMD's hardware? Do you know anything about standardization efforts in DirectX and OpenGL? Do you have any idea on how every new feature is added into these common APIs (of which DirectX isn't an open one BTW which doesn't stop it from supporting different hardware vendors - because this was always the goal of MS in DX development and it clearly isn't in Mantle)?

NV (and ironically AMD themselves) already stated that OpenGL provides ARB extensions (meaning that they're not vendor specific) that allows pretty much the same CPU efficiency that you see in Mantle benchmarks.

MS has recently stated that they are beginning the design phase on DirectX Next.

Now tell me why would the whole industry move from the proven themselves time and time again for years ways of building a new version of common APIs in favor of AMD's Mantle effort which began in such a way as to make everyone but AMD look bad? Why would NV spend time and money on working on Mantle support via a non existent right now API standardization authority instead of continuing their work on OpenGL where everything is set and properly arbitrated for more than 20 years already? Why would MS invest anything at all into Mantle instead of going the usual route for DX12? Especially why would MS do anything in light of AMD's essential backstabbing DX11's multithreading with providing no support for it in their drivers and building Mantle instead?

AMD talk much. They talked about how they support only open standard for years - then they've went and built Mantle which is anything but open. They talked about how they don't do game-specific optimizations for a time - and then went and started doing them as soon as that become in their interest. AMD's proven again and again that they can't be trusted about such things - but somehow there are still a lot of people thinking that AMD is different from other companies. They're all the same, guys. They all lie if that's to their benefit.

How do you know this again?

People know stuff.

 

[BriareosGAF]

So you both applaud their decision not to support the standard mutlithreading features of DirectX11 well, and admit that Mantle development is to the detriment of their driver quality in standard APIs?
Just for the record.

Are you implying that current deferred context support in DX11.x is possible to support well? Just for the record. Nothing the client can do will amortize the fact that the framework/drivers are doing serialization/replay on execute.

 

[artist]

You can't make an open standard from a standard that was made as a closed one, developed in secrecy for two years without any input from other vendors.

Again, why is the onus on AMD when the other vendors stated "sorry cant do" to the devs?

I'm trying to find a definitive source where AMD flat out said that Mantle was an "open standard" and so far havent been able to find one ..

 

[twobear]

There aren't multiple vendors, yet. The plan is to bring more people into the fold. And they've already said that the collaborative effort that comes may not be "Mantle" at all. It doesn't have to follow some cookie cutter scheme to be open.

'Bringing more developers into the fold' is, basically by definition, not open. Open means that the specifications are available to everybody, not the select few that AMD let see and get involved.

If you're talking about 'bringing people in on' something, it's not an open standard. I don't see how this is debatable.

It might become an open standard later, but if it became genuinely open it would completely undermine AMD's only reason for expending money on it, which is a leg-up over nVidia due to the fact that there's GCN-based GPU parts in PS4 and Xbone. Since it would go against AMD's interests to make a genuinely open standard that nVidia can easily implement for similar performance gains, I'm pretty sure 'I'll believe it when I see it' is a pretty reasonable response to the concept of Mantle being open, and not at all FUD.